The explosive interest in GPGPU’s is something that has been a long time coming, and it is only now that there is a real push for it. It could unlock some of that latent potential that’s been building up over the years, along with SSD’s to give systems that bandwidth kick. But there is a real concern that’ll probably show its head soon enough, that of GPGPU viruses.
The internals of the GPU are somewhat out of site, the only way to peer into them is with the use of 3rd party apps, like Rivatuner. But with the interest in GPGPU capabilities, the OS’s will need to provide an easier way for users to monitor their GPU’s, like a modified taskmanager. Think down the line, the GPU will become like that of a specialised CPU, with its own processes and memory to manage. If there was no way to monitor it sufficiently, then what if a virus was to make it’s way onto your GPU? You would have no way to know if it’s consuming your resources, AntiVirus wouldn’t detect it, no way to shut it down, as well as consume large amounts of power should it want to. While a lot of these problems are some ways off, it is important to think about them now and at least prepare for them.
To begin with, OpenCL and CUDA may be like that of API’s with regard to the GPU, not directly executing code but just telling the GPU what to do, these would still require a thread on the OS, to which AV’s could monitor. As the technologies develop, it may then be possible to execute code directly on the GPU, which could bypass the AV, resulting in the above situations. I do not fully understand the capabilities or the processes involved with executing GPU based applications since I am not a programmer, but I still think these are justifiable concerns.
Am I overly paranoid, as well as getting things muddled up?
The internals of the GPU are somewhat out of site, the only way to peer into them is with the use of 3rd party apps, like Rivatuner. But with the interest in GPGPU capabilities, the OS’s will need to provide an easier way for users to monitor their GPU’s, like a modified taskmanager. Think down the line, the GPU will become like that of a specialised CPU, with its own processes and memory to manage. If there was no way to monitor it sufficiently, then what if a virus was to make it’s way onto your GPU? You would have no way to know if it’s consuming your resources, AntiVirus wouldn’t detect it, no way to shut it down, as well as consume large amounts of power should it want to. While a lot of these problems are some ways off, it is important to think about them now and at least prepare for them.
To begin with, OpenCL and CUDA may be like that of API’s with regard to the GPU, not directly executing code but just telling the GPU what to do, these would still require a thread on the OS, to which AV’s could monitor. As the technologies develop, it may then be possible to execute code directly on the GPU, which could bypass the AV, resulting in the above situations. I do not fully understand the capabilities or the processes involved with executing GPU based applications since I am not a programmer, but I still think these are justifiable concerns.
Am I overly paranoid, as well as getting things muddled up?